94th American Meteorological Society Annual Meeting
11th Symposium on the Urban Environment
February 2-6, 2014
Multi-scale Study of Chicago Heat
Island and the Impacts of
Climate Change
P. Conry, A. Sharma, M. Potosnak,
H. J. S. Fernando, and J. Hellmann
Environmental Fluid Dynamics Laboratories
University of Notre Dame
Chicago heat island
• July 1995 heat wave
– Record 106 °F at Chicago Midway
– 465 deaths
• Chicago Climate Action Plan
• Want to better understand Chicago’s vulnerabilities
– Plan and implement UHI adaptation initiatives in face of climate change
Infrared image showing UHI
taken near Chicago lakeshore
Multi-scale approach
• UHI is phenomenon with effects ranging the physical scales
– Need modeling tool to do same
• Multi-model approach via dynamical downscaling
– Global to regional to city to neighborhood scales
– 2.5 degrees
9 km
3 km
1 km
333 m
Global
Community
Climate System
Model (CCSM5)
Mesoscale
Weather Research
and Forecasting (WRF)
model coupled with
sophisticated urban
model
2m
Micro-scale
ENVI-met
Multi-scale approach
1 km
0.333 km
Mesoscale model
• Used sensitivity studies of metropolitan Chicago to test WRF
urban schemes
• WRF model uses BEP+BEM scheme with modified parameters
to model urban climate
• Initial and boundary conditions from either North American
Regional Reanalysis (NARR) dataset (present) or CCSM5 (future)
Micro-scale model
• ENVI-met v3.1 developed by Michael Bruse (Bruse and Fleer 1998)
• 3D Reynolds Averaged Navier-Stokes model
– Boussinesq approximation
– k-𝜀 1.5 order turbulence closure scheme
• 1D model supplies lateral/upper boundary conditions for 3D model
– Only initial conditions fed by user; thereafter marches forward in time without
further nudging
• Includes additional models:
Vegetation
Model
Radiation
Model
Atmospheric
Model
www.envi-met.com
Surface
Boundaries
Soil Model
Micro-scale model
ENVI-met model domain and experimental locations
430 x 438 m2
Experimental campaign
FB
MS1
July 24 to
August 20,
2013
Location
Tower Instruments
McGowan South building
rooftop (MS)
MS1a,b 1 RM Young 81000 sonic anemometer, 1 Campbell
Scientific HMP45C Temp/RH Probe, 3 thermocouples
Munroe Courtyard (MC)
990 Fullerton building
rooftop (FB)
a
MS2b
2 thermocouples
MC1b
2 thermocouples
MC2a
1 thermocouple
FBa
1 RM Young 81000 sonic anemometer, 3
thermocouples
used for model initialization
b used
for model validation
Model validation
• August 17 and 18 – period of relatively steady regional conditions
– Low synoptic gradient winds; local eastern lake breeze dominates
– Expect ENVI-met model to perform well
• 1 AM August 19 – front passage
– Sudden transition to synoptic southwest wind (matches NARR wind direction)
– Expect divergence of ENVI-met model from observations
NCEP NARR 850 mbar
Aug 17
Aug 18
Aug 19
Model validation
• August 18-19 nighttime front
transition affects local conditions
– Vertical air temperature
gradient
– Sensible heat flux (H)
Model validation
• August 17, 2013
• Difference measures (Wilmott
1982)
– Root mean square error (RMSE)
– Mean average error (MAE)
– Index of agreement (d)
• Values approaching 1.0 – good model
performance
Station
RMSE (°C)
MAE (°C)
d
Obs. WRF Obs. WRF Obs. WRF
MS1
1.90 1.29 1.80 1.01 .802 .891
MC1
1.20 0.65 1.09 0.52 .907 .971
MS2
1.54 1.06 1.37 0.72 .855 .921
Model validation
• August 18, 2013
• Difference measures pre-1 AM:
Station
RMSE (°C)
MAE (°C)
d
Obs. WRF Obs. WRF Obs. WRF
MS1
1.96 1.13 1.77 0.97 .709 .863
MC1
1.08 0.99 0.98 0.89 .897 .915
MS2
1.44 1.38 1.12 1.28 .777 .782
• After 1 AM begins to diverge at
MS1 and MC1
– For instance for MS1 WRFinitialized:
• RMSE = 1.60 °C, MAE = 1.59 °C
• d = 0.412
Climate change impacts
• Future conditions
– CCSM5 output averaged over years 2076 to 2081 fed into WRF-urban
– Average over the entire month of August to get an average future
August day – provides initial conditions to ENVI-met model
• Present conditions
– Take August 18 as “typical” day
– Ensemble to obtain current climatology not yet completed (future work)
• For impacts on pedestrian thermal comfort 2 m results are
relevant
Climate change impacts
12:00 LST
Air Temperature Ta
Present
219
219
209
209
199
199
189
189
179
179
169
169
159
159
149
149
139
139
129
129
119
119
Lincoln Park, 08-18-2013,
Test 4 12:00:00 18.08.2013
Future
Pot. Temperature
unter 296.98 K
296.98 bis 297.06 K
Y (m)
Y (m)
x/y cut at z= 0
109
297.06 bis 297.15 K
297.15 bis 297.23 K
109
99
99
89
89
79
79
69
69
59
59
49
49
39
39
29
29
19
19
9
9
297.23 bis 297.31 K
297.31 bis 297.40 K
297.40 bis 297.48 K
297.48 bis 297.57 K
297.57 bis 297.65 K
über 297.65 K
N
0
10
20
30
ΔTa = 0.38 °C
<Left foot>
40
50
60
70
80
90
< 23.8 °C
23.8 °C
23.9 °C
24.0 °C
24.1 °C
100
110
X (m)
120
130
140
150
160
170
24.2 °C
24.3 °C
24.4 °C
24.5 °C
> 24.5 °C
180
190
200
210
0
10
20
ΔTa = 0.60 °C
<Left foot>
30
40
50
60
70
80
90
< 26.5 °C
26.5 °C
26.6 °C
26.7 °C
26.8 °C
<Right foot>
100
110
X (m)
120
130
140
150
160
170
26.9 °C
27.0 °C
27.1 °C
27.2 °C
> 27.2 °C
180
190
200
210
Climate change impacts
Mean radiant temperature (Tmrt) contour
at 12:00 LST on an August 2081 day
LST (hr)
Tmrt (°C)
Ta (°C)
08:00
23-72
23.1-23.6
10:00
30-76
24.7-25.3
12:00
33-81
26.4-27.1
14:00
33-83
27.5-28.2
16:00
31-81
27.6-28.2
18:00
20:00
21-42
16-21
26.6-27.2
25.3-26.0
22:00
15-19
24.6-25.3
00:00
14-18
24.1-24.7
02:00
13-18
23.6-24.3
04:00
13-18
23.3-23.9
06:00
17-45
23.1-23.7
Conclusions
• Multi-model chain coupled via dynamical downscaling –
comprehensive instrument for studying UHI
• ENVI-met model can predict urban microclimate accurately
during periods of steady synoptic conditions
– Allows application to future climate scenarios where change in
average conditions is concern
• Source of initial conditions matters
– Area-representative WRF-urban output outperforms observations
• Micro-scale climate change output being applied to building
energy in addition to pedestrian thermal comfort
Acknowledgements
• This research was funded by NSF Grant No. 0924592 and
Notre Dame Environmental Change Initiative
• Special thanks to DePaul University’s Facility Operations for
access experimental locations
• Special thanks also to Scott Coppersmith and Raffaele Quarta
for aid during experimental campaign
• Finally, the authors gratefully acknowledge Ed Bensman at
Notre Dame’s Engineering and Science Computing for
providing computational resources
Thank you